Train Test Validation Suite

Status	Check	Condition	More Info
✖	Image Property Drift	drift score < 0.2 for image properties drift	Failed for 2 out of 7 properties. Found 2 numeric properties with Kolmogorov-Smirnov above threshold: {'Aspect Ratio': '0.37', 'Area': '0.78'}
✖	Property Label Correlation Change	Train-Test properties' Predictive Power Score difference is less than 0.2	Properties and classes with PPS difference above threshold: {'Brightness': {'Subtype_Ia': '0.33'}, 'RMS Contrast': {'Subtype_Ia': '0.38'}, 'Mean Green Relative Intensity': {'Subtype_Ia': '0.3'}}

Conditions Summary

Status	Condition	More Info
✖	drift score < 0.2 for image properties drift	Failed for 2 out of 7 properties. Found 2 numeric properties with Kolmogorov-Smirnov above threshold: {'Aspect Ratio': '0.37', 'Area': '0.78'}

Conditions Summary

Status	Condition	More Info
✖	Train-Test properties' Predictive Power Score difference is less than 0.2	Properties and classes with PPS difference above threshold: {'Brightness': {'Subtype_Ia': '0.33'}, 'RMS Contrast': {'Subtype_Ia': '0.38'}, 'Mean Green Relative Intensity': {'Subtype_Ia': '0.3'}}

Status	Check	Condition	More Info
✓	Label Drift	categorical drift score < 0.15 and numerical drift score < 0.15	Passed for 1 label properties out of 1 label properties. Found label property "Samples Per Class" has the highest categorical drift score: 0
✓	New Labels	Percentage of new labels in the test set is less or equal to 0.5%	No new labels were found in test set.

Conditions Summary

Status	Condition	More Info
✓	categorical drift score < 0.15 and numerical drift score < 0.15	Passed for 1 label properties out of 1 label properties. Found label property "Samples Per Class" has the highest categorical drift score: 0

Check	Summary
Heatmap Comparison	Check if the average image brightness (or bbox location if applicable) is similar between train and test set. Read More...
Image Dataset Drift	Calculate drift between the entire train and test datasets (based on image properties) using a trained model. Read More...